Why do I get errors about UNIQUE KEYs?

SingleStoreDB does not support unique keys unless the columns in the unique key are a superset of the columns in the shard key.

This requirement is necessary because the shard key informs the query execution system which partition contains the searched value(s). Then the unique key index locates the distinct values searched for within the partition and enforces uniqueness within the partition. For more information about shard keys, see Shard Keys.

Unique Key Restriction Workarounds

  • Have an application enforce the uniqueness of the proposed unique key by confirming the value to be inserted or updated does not exist in the table. Additionally, a non-unique key can be created on the same column to increase execution time.

  • Use the AUTO_INCREMENT argument for one column and use the UNIQUE constraint on the other column. Set the shard key on the column that has the UNIQUE constraint on it:

    CREATE TABLE user_emails(
    email VARCHAR(140) NOT NULL,
    SHARD KEY (email),
    CONSTRAINT emp_email_uk UNIQUE (email)

    The AUTO_INCREMENT column's uniqueness is not enforced directly, but the way the values for it are generated guarantees uniqueness.

  • Create two tables for the same data set, but with different keys to enforce uniqueness on different columns. In the following example, the users table enforces uniqueness and shards on the id column because a primary key acts as both a shard key and a unique index. The users_email table enforces uniqueness and shards on the email column while adding an additional index on the id column.

    A stored procedure can be used to update both tables when a new user is added, so both UNIQUE constraints are enforced.

    /* Table #1 */
    CREATE TABLE users (
    id bigint NOT NULL PRIMARY KEY,
    email varchar(140) NOT NULL,
    /* Table #2 */
    CREATE TABLE users_email (
    id bigint NOT NULL
    email varchar(140) NOT NULL PRIMARY KEY,

Last modified: March 14, 2023

Was this article helpful?